This article is a PERFECT example of why you don't let laymen tech writers do serious benchmarking or technical reviews of OSes and hardware.
There are so many things that would affect the results it would take a longer article to explain how messed up this article is and how it is NOT reflective of anything.
Here is one example to consider, the 'multi-tasking' benchmarks in this article shows OS X to beat Win7, yet as ANY OS ENGINEER will point out the OS X kernel is optimized for multi-tasking on one CPU core, and when it runs on multi-core computers, OS X has this horrible side effect called FUNNEL LOCKS, that essentiall shut down the additional core as applications and the OS isolated threads.
This example BY DEFINITION means that OS X cannot multi-task as well as Win7 on a Dual-Core computer. This is why if you notice one of the applications the author used for rendering, was able to run faster on Win7, because the OS X CPU cores were possibly being locked.
Additionally, the FUNNEL locks of the OS X kernel design means that if you only have ONE active application or two that probably don't force FUNNEL locking of the CPU cores, then multi-tasking will be somewhat OK, yet if you introduct JUST ONE application that creates FUNNEL locks in the OS, the whole system performance not only becomes bad at multi-tasking but can almost seem like a cooperative multi-tasking OS, as applications will start to 'stutter'. And 'stuttering' is not something you will EVER see in this context on a Windows NT OS because of the SMP handling in the kernel design.
In fact, these 'locks' and 'stuttering' doesn't exist in Linux or OpenBSD either, as the FUNNEL locks of OS X are because of queue handling of single CPU multi-tasking that was attempted to be optimized with the MACH/BSD kernel mergers from the original XNU project.
These 'locks' are also why the new multi-core libraries were added to Snow Leopard, but are a bandaid, as it doesn't fix the OS's ability to prevent funnel locks, it just provides an alternative development framework that avoids the funnel locks. i.e. Existing applications and any application not made with the new multi-core libraries will still bring OS X to locking the CPU cores, when Apple should have revisited the inherent flaw of the OS X kernel design that allows funnel locks.
Next we could go on about Quicktime being used for benchmarks to the HORRID drivers supplied with Boot Camp that some believe Apple artificially retards or are just retarded about making them.
Want a real test, go find ANY Vista or Win7 vs OS X test that runs mainstream applications available on both OSes, oh, for example, how about Adobe products - which virtually every test shows Vista and Win7 to be faster than OS X Tiger, Leopard, and Snow Leopard...
Heck even go back to when Vista was 'doggy' before SP1 and optimized drivers:
http://www.broadcastnewsroom.com/articles/viewarticle.jsp?id=167680
Give the 'real' tech world and OS Engineers a few more weeks to iron out Snow Leopard and Win7 and show the world that OS X is only faster when running Apple's own freaking Quicktime - Geesh...
So, let's just leave it here that this article is sad, and painfully sad...
This post was edited by thenetavenger on Friday, October 16, 2009 at 12:00.
|